DonAdam
Explorer
*shrug* I was envisioning the PCs cornering the wife in her room at night and demanding she talk to him or else.
Slow your roll. This is D&D, not FATAL.

I don't have a problem with making some actions inherently counterproductive to certain challenges. Seems logical and obvious from a broad design standpoint.
As for the Duke scenario in particular, I think the approach of fixed contextual DC's and making some skills irrelevant is vastly more realistic than previous editions. Social skills were basically bludgeons to get what you wanted. I had a problem with it being automatic for the NPCs just as I have a problem with it being automatic for the PCs. Persuasion of any kind should be context dependent, so I'm happy to see that its decisive here.
I really hope they get the DC's right. Multiple rolls certainly helps. Given the 3.5 DC's, it was all up to circumstance bonuses anyway. Which meant it was all up to DM fiat. That seemed alot more like railroading.
Also, signaling that you can use any skill within reason works great if you've got an active group of good roleplayers. Two things can break that down:
1) Quiet players. If I want to highlight their character's abilities, I might need to custom tailor a situation in which they can shine. Completely open-ended skills might allow the more active party members (Thor bless them) to steal the spotlight. Creativity is great, but its not the only thing that matters at the table.
2) Disruptive players. I'm sorry, but I just can't see designing a scenario with open-ended skill checks for a one-shot, con or gameday. The average player is just too weird, and in those one-shot scenarios they have hyperbolic discount rates since it's a one-shot. They're going to see what they can get away with. And no, I don't want to deal with that "in game" or after the fact.